| { | |
| "_name_or_path": "./checkpoints/vicuna-7b-v1.5-16k", | |
| "architectures": [ | |
| "STLlamaForCausalLM" | |
| ], | |
| "attention_bias": false, | |
| "bos_token_id": 1, | |
| "eos_token_id": 2, | |
| "freeze_st_mlp_adapter": false, | |
| "st_hidden_size": 64, | |
| "lin_hidden_size": 128, | |
| "time_steps": 12, | |
| "st_select_layer": -2, | |
| "st_tower": "ST_Encoder", | |
| "hidden_act": "silu", | |
| "hidden_size": 4096, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 11008, | |
| "max_position_embeddings": 4096, | |
| "max_sequence_length": 16384, | |
| "model_type": "STLlama", | |
| "num_attention_heads": 32, | |
| "num_hidden_layers": 32, | |
| "num_key_value_heads": 32, | |
| "pad_token_id": 0, | |
| "pretrain_ST_model_path": "./checkpoints/st_encoder/pretrain_stencoder.pth", | |
| "pretraining_tp": 1, | |
| "rms_norm_eps": 1e-05, | |
| "rope_scaling": { | |
| "factor": 4.0, | |
| "type": "linear" | |
| }, | |
| "rope_theta": 10000.0, | |
| "sep_st_conv_front": false, | |
| "tie_word_embeddings": false, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.34.0", | |
| "tune_st_mlp_adapter": true, | |
| "use_cache": false, | |
| "use_st_proj": true, | |
| "use_st_start_end": true, | |
| "vocab_size": 32003 | |
| } | |